## [1] '3.10.3'

## virtualenv: r-reticulate

## Using virtual environment "C:/Users/abe.y/Documents/.virtualenvs/r-reticulate" ...

2.5 input定義

## Total Observations: 366 (days)
## Input Table Columns (8):
##   Date: datetime
##   Dependent: revenue [revenue]
##   Paid Media: Net.Spend, Tv.Spend
##   Paid Media Spend: Net.Spend, Tv.Spend
##   Context: temperature, rain, Weekend.FLG
##   Organic: 
##   Prophet (Auto-generated): trend, season on US
##   Unused variables: calculation_1
## 
## Date Range: 2019-04-01:2020-03-31
## Model Window: 2019-04-01:2020-03-31 (366 days)
## With Calibration: FALSE
## Custom parameters: None
## 
## Adstock: geometric
## Hyper-parameters: [0;31mNot set yet[0m

2.5.2 ハイパーパラメータ指定

## [1] "Net.Spend_alphas" "Net.Spend_gammas" "Net.Spend_thetas" "Tv.Spend_alphas" 
## [5] "Tv.Spend_gammas"  "Tv.Spend_thetas"

2.5.3 input

## Total Observations: 366 (days)
## Input Table Columns (8):
##   Date: datetime
##   Dependent: revenue [revenue]
##   Paid Media: Net.Spend, Tv.Spend
##   Paid Media Spend: Net.Spend, Tv.Spend
##   Context: temperature, rain, Weekend.FLG
##   Organic: 
##   Prophet (Auto-generated): trend, season on US
##   Unused variables: calculation_1
## 
## Date Range: 2019-04-01:2020-03-31
## Model Window: 2019-04-01:2020-03-31 (366 days)
## With Calibration: FALSE
## Custom parameters: None
## 
## Adstock: geometric
## Hyper-parameters ranges:
##   Net.Spend_alphas: [0.5, 3]
##   Net.Spend_gammas: [0.3, 1]
##   Net.Spend_thetas: [0, 0.3]
##   Tv.Spend_alphas: [0.5, 3]
##   Tv.Spend_gammas: [0.3, 1]
##   Tv.Spend_thetas: [0.1, 0.4]
##   train_size: [0.7]

2.5.4 モデル構築

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |========================                                              |  35%
  |                                                                            
  |=================================================                     |  70%
 
##   Finished in 0.05 mins
## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |========================                                              |  35%
  |                                                                            
  |=================================================                     |  70%
 
##   Finished in 0.04 mins
## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |========================                                              |  35%
  |                                                                            
  |=================================================                     |  70%
 
##   Finished in 0.04 mins
## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |========================                                              |  35%
  |                                                                            
  |=================================================                     |  70%
 
##   Finished in 0.04 mins
## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |========================                                              |  35%
  |                                                                            
  |=================================================                     |  70%
 
##   Finished in 0.05 mins

## Total trials: 5
## Iterations per trial: 20 (21 real)
## Runtime (minutes): 0.37
## Cores: 7
## 
## Updated Hyper-parameters:
##   Net.Spend_alphas: [0.5, 3]
##   Net.Spend_gammas: [0.3, 1]
##   Net.Spend_thetas: [0, 0.3]
##   Tv.Spend_alphas: [0.5, 3]
##   Tv.Spend_gammas: [0.3, 1]
##   Tv.Spend_thetas: [0.1, 0.4]
##   lambda: [0, 1]
##   train_size: [0.7]
## 
## Nevergrad Algo: TwoPointsDE
## Intercept: TRUE
## Intercept sign: non_negative
## Time-series validation: TRUE
## Penalty factor: FALSE
## Refresh: FALSE
## 
## Convergence on last quantile (iters 20:21):
##   DECOMP.RSSD NOT converged: sd@qt.20 0.065 > 0.052 & |med@qt.20| 0.35 > 0.3
##   NRMSE NOT converged: sd@qt.20 0.25 > 0.087 & |med@qt.20| 6 > 5.8

クロスバリデーションによる誤差の確認

パレート最適化の計算

## 
 00:00:00 [=========================================] 100% | 1

## Plot Folder: C:/Users/abe.y/.exploratory/projects/RobynDemo_bZg2DkZ1/markdown_output/Robyn_202307281012_init/
## Calibration Constraint: 0.1
## Hyper-parameters fixed: FALSE
## Pareto-front (1) All solutions (1): 1_2_1

2.6 モデルの選択

2.7 予算の最適化

## Exported directory: C:/Users/abe.y/.exploratory/projects/RobynDemo_bZg2DkZ1/markdown_output/Robyn_202307281012_init/
## Exported model: 1_2_1
## Window: 2019-04-01 to 2020-03-31 (366 days)
## Time Series Validation: TRUE (train size = 70%)
## 
## Model's Performance and Errors:
##     Adj.R2 (test): -2.641 | NRMSE = 5.427 | DECOMP.RSSD = 0.2062 | MAPE = 0
## 
## Summary Values on Selected Model:
##      variable    coef decompPer decompAgg   ROI mean_response mean_spend
## 1 (Intercept) 358.54K    85.97%   131.23M     -             -          -
## 2       trend   0.098     7.79%   11.888M     -             -          -
## 3      season   0.062     0.00%   -5.387K     -             -          -
## 4 temperature  1.126K     4.51%     6.88M     -             -          -
## 5        rain -50.033    -0.06%  -99.041K     -             -          -
## 6 Weekend.FLG   0.074     0.48%   729.76K     -             -          -
## 7   Net.Spend  22.06K     0.41%   626.92K 0.058        215.23    29.635K
## 8    Tv.Spend 21.634K     0.91%    1.388M 0.025        4.633K     149.6K
## 
## Hyper-parameters:
##     Adstock: geometric
##     channel    alphas    gammas    thetas
## 1 Net.Spend 2.6644425 0.6383240 0.1065636
## 2  Tv.Spend 0.7541675 0.7713408 0.1923280

## [1] "Net.Spend" "Tv.Spend"

## Model ID: 1_2_1
## Scenario: max_response
## Use case: total_metric_default_range + historical_budget
## Window: 2020-03-02:2020-03-31 (30 days)
## 
## Dep. Variable Type: revenue
## Media Skipped: 
## Relative Spend Increase: -0% (0)
## Total Response Increase (Optimized): 2.84%
## 
## Allocation Summary:
##   
## - Net.Spend:
##   Optimizable bound: [-30%, 20%],
##   Initial spend share: 15.8% -> Optimized bounded: 11.1%
##   Initial response share: 1.11% -> Optimized bounded: 0.418%
##   Initial abs. mean spend: 14.91K -> Optimized: 10.44K [Delta = -30.0%]
##   
## - Tv.Spend:
##   Optimizable bound: [-30%, 20%],
##   Initial spend share: 84.2% -> Optimized bounded: 88.9%
##   Initial response share: 98.9% -> Optimized bounded: 99.6%
##   Initial abs. mean spend: 79.53K -> Optimized: 84K [Delta = 5.6%]